We present the first Audio-Visual+ Emotion recognition Challenge and workshop (AV+EC 2015) aimed at comparison of multimedia processing machine learning methods for automatic audio, visual physiological emotion analysis. This is 5th event in AVEC series, but very that bridges across video data. The goal to provide a common benchmark test set multimodal information bring together communities, compare relative merits three approaches under well-defined strictly comparable conditions establish...

Spontaneous facial expression recognition under uncontrolled conditions is a hard task. It depends on multiple factors including shape, appearance and dynamics of the features, all which are adversely affected by environmental noise low intensity signals typical such conditions. In this work, we present novel approach to Facial Action Unit detection using combination Convolutional Bi-directional Long Short-Term Memory Neural Networks (CNN-BLSTM), jointly learns in deep learning manner....

Depression is a serious mental disorder affecting millions of people all over the world. Traditional clinical diagnosis methods are subjective, complicated and require extensive participation clinicians. Recent advances in automatic depression analysis systems promise future where these shortcomings addressed by objective, repeatable, readily available diagnostic tools to aid health professionals their work. Yet there remain number barriers development such tools. One barrier that existing...

Attention Deficit Hyperactivity Disorder (ADHD) and Autism Spectrum (ASD) are neurodevelopmental conditions which impact on a significant number of children adults. Currently, the diagnosis such disorders is done by experts who employ standard questionnaires look for certain behavioural markers through manual observation. Such methods their not only subjective, difficult to repeat, costly but also extremely time consuming. In this work, we present novel methodology aid diagnostic predictions...

This article proposes to recognise the true (self-reported) personality traits from target subject's cognition simulated facial reactions. approach builds on following two findings in cognitive science: (i) human partially determines expressed behaviour and is directly linked traits; (ii) dyadic interactions, individuals' nonverbal behaviours are influenced by their conversational partner's behaviours. In this context, we hypothesise that during a interaction, reactions driven main factors:...

Current vision based approaches for automatic prediction of mental health conditions like depression and anxiety, rely on models that use behavioural features (usually extracted from faces) only do not take personality into account. However, there is a considerable amount evidence people with certain traits are more prone to anxiety disorders. In order exploit the underlying relationship between these conditions, we propose combination consisting observed facial behaviour self-reported...

This article aims to solve two important issues that frequently occur in existing automatic personality analysis systems: 1. Attempting use very short video segments or even single frames, rather than long-term behaviour, infer traits; 2. Lack of methods encode person-specific facial dynamics for recognition. To deal with these issues, this paper first proposes a novel Rank Loss which utilizes the natural temporal evolution actions, labels, self-supervised learning dynamics. Our approach...

Recent research shows that in dyadic and group interactions individuals' nonverbal behaviours are influenced by the of their conversational partner(s). Therefore, this work we hypothesise during a interaction, target subject's facial reactions driven two main factors: (i) internal (person-specific) cognition, (ii) externalised partner. Subsequently, our novel proposition is to simulate represent (i.e., listener) cognitive process form person-specific CNN architecture whose input audio-visual...

Self-report questionnaires like PHQ-9 and GAD-7 are often used in the field of psychology to detect mental health problems measure their severity. Their validity has been well established by a number previous studies. However, most studies have self-administration method study validity. In context automating administering as part an interaction scenario led virtual human, we investigate if human administration these can be considered equivalent (through electronic form). Additionally, also...

Facial landmark detection in real world images is a difficult problem due to the high degree of variation pose, facial expression and illumination, presence occlusions background clutter. We propose system that addresses head pose expressions guided unsupervised learning approach establish mode specific models. To detect 68 fiducial points we employ Local Evidence Aggregated Regression, which local patches provide evidence location target point using Support Vector Regressors. improve an...

This paper introduces FugaciousFilm, a soap film based touch display, as platform for Attentive Interaction that encourages the user to be highly focused throughout use of interface. Previous work on ephemeral interfaces has primarily development ambient and peripheral displays. In contrast, FugaciousFilm is an display aims promote attentive interaction. We present iterative process developing this interface, spanning technical explorations, prototyping study. report lessons learnt when...

Current approaches to automatic analysis of facial action units (AU) can differ in the way face appearance is represented. Some works represent whole face, dividing bounding box region a regular grid, and applying feature descriptor each subpatch. Alternatively, it also common consider local patches around landmarks, apply descriptors them. Almost invariably, all features from these are combined into single vector, which input learning routine inference. This constitutes so-called...

Learning regression-based machine learning models for computer vision problems is a challenging task due to noisy features, variation in pose and illumination, occlusion, etc. Typically the problem compounded by non-uniform distribution of labels training data, resulting parts label space that suffer from data sparsity imbalance general. Deep Convolutional Neural Networks (CNN) have shown remarkable success on number tasks such as object classification face recognition. However, they too...

The Boltzmann equation, a six-dimensional integro-differential governs the fluid flow behavior at molecular level for wide range of physical phenomena, including shocks, turbulence, diffusion, and non-equilibrium chemistry which are beyond reach continuum modelling based on Navier-Stokes equations. Despite equation's applicability, its deterministic solution presents huge computational challenge, has been so far tractable only in simplified forms. We implement Discontinuous Galerkin Fast...

We have studied in-plane anisotropic magnetoresistance (AMR) in cobalt films with overlayers having designed electrically interface transparency. With an opaque cobalt/overlayer interface, the AMR ratio is shown to vary inverse proportion film thickness; indication that a consequence of scattering both volume and interfacial contributions. The anisotropy opposes contribution, causing diminish as thickness reduced. An intrinsic effect explains significantly reduced ultra-thin films.

Virtual human technologies are now being widely explored as therapy tools for mental health disorders including depression and anxiety. These leverage the ability of virtual agents to engage in naturalistic social interactions with a user elicit behavioural expressions which indicative Research efforts have focused on optimising human-like expressive capabilities human, but less attention has been given investigating effect mediation expressivity user. In addition, it is still not clear what...

When the flow is sufficiently rarefied, a temperature gradient, for example, between two walls separated by few mean free paths, induces gas flow---an observation attributed to thermo-stress convection effects at microscale. The dynamics of overall process governed Boltzmann equation---an integro-differential equation describing evolution molecular distribution function in six-dimensional phase space---which models dilute behavior level accurately describe wide range phenomena. Approaches...

This paper presents GPU parallelization for a computational fluid dynamics solver which works on mesh consisting of polyhedral cells, where each cell has an arbitrary number faces and face vertices. The is achieved using NVIDIAs compute unified device architecture (CUDA). developed code specifically targets performance improvement NVIDIA Tesla accelerator GPUs. implementation been carried out in general purpose open-source CFD framework namely OpenFOAM capable solving flow problems involving...

The direct simulation Monte Carlo (DSMC) method takes advantage of phenomenological models that can efficiently reproduce macroscopic transport and chemistry rates from elementary collisional kinetic data. As in classical theory gases, reproduction viscosity, thermal conductivity, diffusion coefficients rely on the scattering dynamics particles undergo during collisions. In this work, still under-explored Kersch & Morokoff's (M-1) model is implemented DSMC-SPARTA solver. essence, a...